ARTICLE IN PRESS

JOURNAL OF
SOUND AND
VIBRATION
Journal of Sound and Vibration 300 (2007) 101–116
www.elsevier.com/locate/jsvi

Optimisation of a resonance changer to minimise the vibration

transmission in marine vessels
Paul G. Dylejkoa,, Nicole J. Kessissogloua, Yan Tsob, Chris J. Norwoodb
a
School of Mechanical and Manufacturing Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
b
Defence Science and Technology Organisation (DSTO), Maritime Platforms Division, Fishermans Bend, Vic. 3207, Australia
Received 2 December 2005; received in revised form 7 June 2006; accepted 24 July 2006
Available online 27 October 2006

Abstract

Structure-borne noise generated by marine vessels is an area that receives much research attention. Significant noise
levels are generated due to onboard machinery such as the diesel engines, gearboxes, generators, and auxiliary machinery.
In the case of ships and submarines, a major source of the radiated noise at low frequencies is due to excitation of the hull
modes resulting from vibration transmission through the propeller-shafting system. Oscillations occur at the propeller due
to small variations in thrust when the propeller blades rotate through the non-uniform wake, resulting in axial excitation of
the propeller at the blade pass frequency. This problem can be addressed by the use of a resonance changer (RC) which
performs the task of a hydraulic dynamic vibration absorber, thereby reducing the vibration transmission and avoiding
excitation of hull axial resonances. This research is concerned with optimisation of both single and dual RC configurations
in a submarine. An optimisation scheme involving a genetic and a general nonlinear constrained algorithm is used to
minimise two fitness functions associated with the vibration transmission to the hull over a low-frequency range. The
dynamic response of the propeller-shafting system is characterised using the transmission matrix approach. This modular
description enables greater flexibility for dynamic modelling of the propeller-shafting system, and can be easily
manipulated for future design modifications.
r 2006 Elsevier Ltd. All rights reserved.

1. Introduction

It is desirable to minimise the vibro-acoustic responses of maritime vessels to improve passenger comfort,
minimise crew fatigue, and in the case of naval vessels, to avoid detection. Oscillations of onboard equipment
such as the diesel engines and the propulsion system generates significant vibration that propagates through
the supporting structure to the hull, where it is radiated as structure-borne noise. The vibration transmission
through the propeller-shafting system of ships and submarines represents a critical issue that must be
addressed in order to reduce the low-frequency acoustic signature. The propeller-shafting system can be
simplified into the key features shown in Fig. 1. Axial excitation of the propeller occurs at low frequencies due
to the non-uniform wake velocity caused by asymmetry in the hull or protrusions of control surfaces. The

Corresponding author. Tel.: +61 3 9626 8218; fax: +61 3 9626 8373.
E-mail address: paul.dylejko@dsto.defence.gov.au (P.G. Dylejko).

0022-460X/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jsv.2006.07.039
 ARTICLE IN PRESS
102 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

Supporting foundation

Journal bearing Flexible

coupling

Excitation

Journal bearing
Thrust
Propeller bearing

Fig. 1. Simplified representation of the propeller-shafting system [1].

oscillations which occur at the propeller are the result of small variations in thrust when the propeller blades
rotate through the non-uniform wake. The frequency of these oscillations is at the blade pass frequency
(rotational speed of the shaft multiplied by the number of blades on the propeller). The disturbances at the
propeller result in vibration transmission through the propeller-shafting system and subsequent axial
excitation of the hull. The engine also introduces excitation to the propeller-shafting system; however, it is
assumed that this is essentially isolated by the flexible coupling which limits the influence of the engine side
from the propeller side of the propulsion system [1]. Development of propeller-shafting models for maritime
vessels has been undertaken by numerous researchers [1–3]. In most of these studies, the aim has been to
reduce the axial vibration and its transmission into the hull. A detailed paper by Goodwin [2] examined the
reduction of excessive vibration through the propeller-shafting system by using an existing hydraulic device
called the ‘‘Michell Thrustmeter’’. This device is located in series between the thrust bearing and supporting
foundation and is used to measure the thrust which is transmitted to the vessel from the propeller-shafting
system. Goodwin adapted and redesigned the Michell Thrustmeter to reduce the vibration transmission
through the propeller-shafting system. For this new application, the device was known as a resonance changer
(RC), also known as a detuner. The RC introduces virtual elastic, damping and inertial influences by hydraulic
means, thereby acting as a dynamic vibration absorber (DVA). The simplified model of the RC introduced by
Goodwin is shown in Fig. 2. The RC consists of a piston of cross-sectional area A0, an oil reservoir of volume
V1 and a pipe connecting these two elements of length L1 and cross-sectional area A1. Goodwin examined the
reduction of the force transmissibility using a simplified spring–mass model of the propeller-shafting system.
Reduction of the force transmissibility was achieved by firstly tuning the natural frequency of the RC to that
of the propeller-shafting system’s natural frequency and then optimising the RC’s damping rate. Goodwin
also offered insight into design values for the RC’s mass and stiffness. The performance of the RC during sea
trials exceeded expectations, eliminating the resonance conditions suffered previously. Although the paper by
Goodwin is comprehensive in all practical respects, a relatively simple model of the propeller-shafting system
was used and the coupling with the hull was not included.
A RC is essentially a hydraulic DVA, where a DVA is a spring–mass–damper device attached to a
‘‘primary’’ physical system with the aim of reducing its response. Frahm is acknowledged to have invented the
DVA in 1911 [4]. Early work on DVAs concentrated on tuning the undamped DVA to the frequency of the
harmonic excitation acting on the primary system. This type of absorber exhibits good performance over a
narrow frequency bandwidth; however, it becomes inefficient as the disturbing frequency shifts. The addition
of damping alleviates the narrow band characteristics of the undamped DVA. DVAs have been utilised
extensively to reduce vibrations in various types of machinery and structures [5–12]. Much work has been
completed on the optimisation of DVAs attached to different structures. Hartog [7] analytically investigated
the design of DVAs for use with single degree-of-freedom (dof) systems by obtaining optimal values of the
absorber spring, mass and damping constants. This analysis however was restricted to primary systems that
did not exhibit damping.
A simple method that has been used to optimise the response of a primary system with an attached DVA is
the single-dof method. The response of a single resonance is minimised in this method by ignoring the response
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 103

vf L1

A0 V1
vr
A1

Fig. 2. Simplified model of the resonance changer [2].

outside the influence of the chosen resonance. The application of DVAs is not restricted to lumped parameter
systems, Neubert [8] used the single-dof method technique to reduce the axial response of a bar at a chosen
resonance frequency. Mobility was used to characterise the dynamic response of the structure. A method for
the efficient reduction of the dynamic response of large-scale multi-dof systems subject to harmonic excitation
using multiple DVAs was reported by Kitis et al. [9]. The reduction in computational expense reported in their
work was achieved by using a re-analysis technique for computing the cost function and its derivatives. The
benefit associated with this technique is that the complete structural analysis is performed only once.
Subsequent calculation of the cost function and its derivatives only require the solution of a set of linear
algebraic equations. The vibratory response over a frequency range of a cantilever beam approximated by 22-
dof with two attached DVAs was then minimised. While the majority of research has been concerned with
examining the dynamic response in the frequency domain, DVAs have been utilised to improve the response in
the time domain as well. Lau and Lam [10] examined both time and frequency domain response optimisation
using four different methods for a single-dof system with primary damping. These methods included the equal
peak method, the minimal variance method, the energy method and the area ratio method. It was found that
time domain optimisation was best suited for reduction of the transient response. Rice [11] examined the use of
multiple DVAs to reduce the maximum response of a continuous system over a frequency range. The primary
system was characterised using modal data. A numerical optimisation of the parameters and positioning of the
multiple DVAs was performed using the SIMPLEX algorithm. Work by Rade and Steffen [12] involved
solving a general nonlinear constrained optimisation problem to obtain the DVA parameters which resulted in
the minimum response of a continuous system over multiple frequency bands. A substructure coupling
technique was used to characterise the frequency response function of the combined system. The influence of
perturbations in the DVAs parameters was also examined.
In this paper, the transmission matrix method, also known as the four-pole parameter method is used to
model the low-frequency dynamic response of the propeller-shafting system in a submarine. Two fitness
functions associated with the vibration transmission to the hull over a low-frequency range as a function of the
RC parameters are developed. Both fitness functions are minimised using a genetic and a general nonlinear
constrained algorithm within realistic constraints, resulting in optimal values for the virtual RC parameters.
The results obtained using the different fitness functions are presented for both single and dual RC
configurations. The subsequent advantages of the different fitness functions and number of RCs are discussed.
A perturbation analysis on the RC parameters is also presented and the effects on the RC performance are
observed.

2. Transmission matrix model of the dynamic system

Due to the symmetry of the geometry and loading of the propeller-shafting system in a submarine, the
transmission matrix method [13,14] has been chosen to characterise the dynamic response. A transmission
matrix schematic of the propeller-shafting system is given in Fig. 3. The mechanical components have been
broken down into subsystems to enable a modular description of the complete system’s dynamic response. The
proposed dynamic model assumes that the propeller and the entrained water around the propeller are
represented as a lumped mass of mass mp with viscous damping cp. The propeller is attached to a continuous
model of the shaft consisting of cross-sectional area As, Young’s modulus Es and density rs . Since the response
at a point along the shaft corresponding to the location of the thrust bearing is desired, an effective length l se is
 ARTICLE IN PRESS
104 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

Fig. 3. Transmission matrix representation of propeller-shafting model connected to the submarine hull.

defined. The thrust bearing is represented by a linear stiffness kb , damping coefficient cb and mass mb . The RC
exhibits inertial, elastic and damping properties, represented by mr , kr and cr , respectively. Referring to Fig. 2,
these virtual mass, stiffness and damping parameters can be expressed by [2]
r1 A20 L1
mr ¼ , (1)
A1

A20 B1
kr ¼ , (2)
V1

A20
cr ¼ 8pm1 L1 . (3)
A21
These parameters are associated with the force required to overcome the inertia of the oil in the pipe,
compress the oil in the reservoir and overcome the viscous resistance of the oil in the pipe. r1 , m1 and B1 are,
respectively, the density, viscosity and bulk modulus of the oil contained within the RC system. The thrust
block is coupled to the hull via a truncated conical shell foundation. E f , rf , nf , hf , respectively, represent
Young’s modulus, density, Poisson’s ratio and thickness of the foundation. The radii of the major and minor
base of the conical shell are a and b, f is the semivertex angle. E h , rh , nh , ro , hh , Ah and l h represent Young’s
modulus, density, Poisson’s ratio, outside radius, thickness, cross-sectional area and length of the cylindrical
hull, respectively. The velocities at the propeller, shaft, thrust bearing, RC, foundation and hull are described
by vp , vs , vb , vr , vf and vh , respectively, while the forces at the previous locations are given by f p , f s , f b , f r , ff
and fh.

2.1. Propeller-shafting system

The transmission matrix parameters of the propeller (ignoring the damping due to the surrounding fluid)
and shaft are, respectively, given by
 
1 jomp
ap ¼ , (4)
0 1
2 3
As E s ks sin ks l s
6 cos ks l se j
6 o cos ks ðl s  l se Þ 7
7
as ¼ 6 7. (5)
4 cos ks ðl s  l se Þ  cos ks l s cos ks l se cos ks l s 5
jo
As E s ks sin ks l s cos ks ðl s  l se Þ
The shaft transmission matrix parameters were obtained by manipulating the receptance matrix for a
free–free rod undergoing longitudinal vibration [15]. ks ¼ o=cLs is the longitudinal wavenumber, and cLs ¼
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 105
pffiffiffiffiffiffiffiffiffiffiffiffi
E s =rs is the longitudinal wave speed of the shaft. The transmission matrix parameters of the thrust bearing
and RC are respectively expressed as
2 3
o2 mb
6 1  jom b7
6 kb þ jocb 7
ab ¼ 6 7, (6)
4 jo 5
1
kb þ jocb
2 3
1 0
ar ¼ 4 jo 5
1 . (7)
kr þ jocr  o2 mr

2.2. Foundation

A simplified model of a truncated conical shell has been used to model the foundation of the propeller-
shafting system in a submarine. It is assumed that the axisymmetric response of the foundation in the low-
frequency range can be approximated using membrane theory. The governing equations can be written as [16]
U ¼ cos fðcos2 f  O2 x2 Þ1
½ðnf tan f=2pE f hf ÞF  ð1  O2 x2 Þu, ð8Þ

N y ¼ ðE f hf =aÞðO2 x= sin f cos fÞ½u cos f þ U, (9)

dF =dx ¼ 2pE f hf O2 xU cos f, (10)

du=dx ¼ ðE f hf sin fÞ1 ½F =2px cos f þ naN y . (11)

The geometric parameters and coordinate system of the truncated conical shell used in Eqs. (8) to (11) are
shown in Fig. 4. U is the axial displacement of a shell cross section, N y is the circumferential stress resultant, u
is the displacement of the shell in the meridional direction and F is the total axial force transmitted through a
shell cross section. The dimensionless coordinate of the conical shell is: x ¼ s=s1 , where s is the meridional
coordinate of the conical shell (distance measured from apex) and s1 is the meridional distance from the apex
to the major base. The completeness parameter is defined as g ¼ b=a. The dimensionless frequency parameter
O is given by O ¼ o=o0 where o20 ¼ E f =rf a2 .

s2

s1 a

s,u b


Fig. 4. Coordinate system used with the conical shell model of the foundation.
 ARTICLE IN PRESS
106 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

The four-pole parameters can be found by numerical integration of the second-order equations for two sets
of initial conditions at the minor base of the conical shell (x ¼ g), corresponding to ðF ; UÞ ¼ ð1; 0Þ and (0, 1).
Using these initial conditions and integrating the differential equations to the major base of the conical shell
(x ¼ 1), results in two sets of influence coefficients, (c11 , c21 ) and (c12 , c22 ). The transmission matrix of the
foundation is then given by
" #
c22 ic 12 =o
af ¼ . (12)
ioc21 c11

It is important to note that the geometry of the conical shell used by Hu and Kana [16] has been inverted to
correspond to the geometry of the foundation shown in Fig. 4.

2.3. Hull

As an initial approximation, the submarine hull was modelled as a one-dimensional rod undergoing axial
excitation. It should be noted that this description does not include radial motion of the hull or the subsequent
effects of radiation loading. Lumped masses have been added to both ends of the rod to represent the main
ballast tank and casing. The transmission matrix parameters associated with the lumped masses and the
cylindrical hull are, respectively, given by
 
1 joml
al ¼ , (13)
0 1
2 3
Ah E h k h
6 cos kh l h j sin kh l h 7
o
ah ¼ 6
4 o
7.
5 (14)
j sin kh l h cos kh l h
Ah E h kh
ml represents the lumped masses
pffiffiffiffiffiffiffiffiffiffiffiffi
ffi added to the fore and aft of the hull, kh ¼ o=cLh is the longitudinal
wavenumber and cLh ¼ E h =rh is the longitudinal wave speed of the hull. rh is the density of the hull adjusted
using Archimedes principle to maintain neutral buoyancy and is defined as
pr2o l h rw  2ml
rh ¼ , (15)
Ah l h
where rw is the density of the surrounding fluid.
The complete hull response bh is the product of the transmission matrices associated with the cylindrical rod
and lumped masses:
bh ¼ al ah al . (16)
The resulting axial driving point impedance subject to a free end can be described by [13]
bh12
Zd ¼ , (17)
bh22
where bh12 and bh22 , respectively, represent the second element in the first and second rows of the matrix bh .

2.4. Force and power transmission through the propeller-shafting system

For N number of RCs in series, the combined response of the complete propeller-shafting system bps is given
by the forward matrix multiplication of the respective four-pole parameters of the subsystems:
Y
N
bps ¼ ap as ab arq af . (18)
q¼1
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 107

The force at the hull resulting from a unit load at the propeller (f p ¼ 1) is defined by [13]
 1
ps bps
12
f h ¼ b11 þ . (19)
Zd

bps ps
11 and b12 represent the first and second elements in the first row of the matrix b . The force
ps

transmissibility T F through the propeller-shafting system is given by T F ¼ jf h j.

The power transmission to the hull has also been examined due to the fact that the hull velocity has a closer
direct correlation to the radiated sound pressure. The velocity at the hull arising from a unit load at the
propeller is defined by

vh ¼ ðbps ps 1
11 Z d þ b12 Þ . (20)
The time-averaged power transmission to the hull resulting from a unitary harmonic load at the propeller
can be expressed as [17]
hPi ¼ 12Reðf nh vh Þ, (21)
where the brackets hi represent time averaged and f nh is the complex conjugate of f h .

3. Optimisation of the RC parameters

3.1. Development of fitness criteria

The force which acts on the propeller in a marine vessel has been shown to be approximately pro-
portional to the propeller rotational speed squared [1,2]. This relationship has been accounted for by
weighting the force transmissibility and time-averaged power transmission by the frequency ratio oi/Do
squared and raised to the fourth power respectively. oi is the discrete frequency in the frequency band
of interest and Do is the frequency bandwidth used in the optimisation. The weighted force transmissi-
bility and weighted time-averaged power transmission at the ith discrete frequency can be, respectively,
expressed as
 o 2
i
T F ;W ðoi Þ ¼ T F ðoi Þ, (22)
Do
 o 4
i
hPW ðoi Þi ¼ hPðoi Þi. (23)
Do
It should be noted that these terms do not represent physical quantities. They are, however, suitable for
measuring the reduction in force transmissibility or time-averaged power transmission. The fitness criteria to
be minimised are the maximum value of the weighted force transmissibility (J T F ;W ) and the maximum value of
the weighted time-averaged power transmission (J hPW i ) and are given by

J T F ;W ðxÞ ¼ 20 log10 max T F ;W ðx; oi Þ , (24)

ol poi pou

J hPW i ðxÞ ¼ 10 log10 max hPW ðx; oi Þi . (25)

ol poi pou

The frequency range included in Eqs. (24) and (25) is bound by lower (ol ) and upper (ou ) limits. x
is a vector containing the virtual mass, stiffness and damping parameters associated with the N RCs and is
given by:

x ¼ f kr;1 mr;1 cr;1 kr;2 mr;2 cr;2 ... kr;N mr;N cr;N gT . (26)
 ARTICLE IN PRESS
108 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

3.2. Genetic algorithm-based optimisation

Only the low-frequency range (o100 Hz) is of interest due to the excitation of the propeller occurring at the
blade pass frequency. Lower (xl ) and upper (xu ) limits are also enforced on the RC parameters, that is,
xl pxpxu . These two conditions require the use of a constrained optimisation technique. There are numerous
optimisation techniques available to solve the constrained fitness criteria defined in the previous section. Many
of these techniques are robust but sometimes fail to find the global optimum. Complex fitness functions such
as those defined in Eqs. (24) and (25) may contain several local optima. Since a mathematical condition
defining the best solution for these cases does not exist, finding this global optimum is usually more
computationally involved. One method that has been frequently used by researchers to solve various nonlinear
optimisation problems is the genetic algorithm or GA [18,19]. GAs are an artificial application of Darwin’s
notion of natural selection and evolution, and have been proven to provide robust and accurate solutions to
these problems. While equality constraints can be included in the system model, inequality constraints cannot
be directly enforced by the GA. The simplest way of applying constraints is by eliminating strings which
violate any of the constraints. This method is not ideal since no information is obtained from infeasible
solutions. In order to overcome this problem, the penalty method can be used. This method reduces the fitness
of the string relative to the constraint violation. Given the constrained optimisation problem: minimise JðxÞ,
subject to hp ðxÞ  0 for p ¼ 1; 2; :::; n where n is the number of constraints, the penalty method transforms the
problem into an unconstrained problem suitable for solution by the GA [19]:
X
n
minimise JðxÞ þ F½hp ðxÞ, (27)
p¼1

where F is the penalty function which is equal to zero when no constraints are violated. A simple penalty
function that has been used extensively is the square of the constraint violation multiplied by a penalty
coefficient (P) and is given by [19]: F½hp ðxÞ ¼ Ph2p ðxÞ. A dynamic penalty function presented by Erbatur et al.
[20] has been used, which adjusts the intensity of the penalty parameter with increasing generations:
F½hp ðxÞ ¼ PG 2 hp ðxÞ. G represents the generation count. In order to provide a fair penalty distribution, the
constraints have been normalised by using
8 xl;p  xp
>
> ; xp oxl;p ;
>
< xm;p
>
hp ðxp Þ ¼ xp  xu;p ; xp 4xu;p , (28)
>
> xm;p
>
>
:
0; xl;p pxp pxu;p ;

where xm;p ¼ ðxl;p þ xu;p Þ=2. The index p in this case represents the pth element in the respective vector. This
normalisation results in three constraints for each RC. Combining this method and the penalty function given
above, the costP functions defined in Eqs. (24) and (25)10are penalised by adding the following term:
Penalty ¼ PG 2 3N p¼1 hp ðxp Þ. Values for P of 2.5 and 2:5  10 have been, respectively, used when considering
the force transmissibility and time-averaged power transmission [20].

4. Results

An optimisation scheme utilising a genetic and a general nonlinear constrained algorithm has been used to
minimise the fitness criteria defined in Eqs. (24) and (25). The genetic algorithm was used to approximately
find the global optima, while the general nonlinear constrained algorithm improved the accuracy of the
approximate solution found by the GA. The physical values associated with the propeller-shafting system,
foundation and hull used in the modelling are presented in Tables 1–3, respectively. The lumped masses added
to the fore and aft of the cylindrical hull were 200 tonnes. The density of the surrounding fluid is assumed
to be 1000 kg/m3. The four-pole parameters of the conical shell obtained from the numerical integration of
Eqs. (8)–(11) are given in Fig. 5. Hysteretic damping was included in the shaft and hull by using a complex
Young’s modulus Eð1 þ jZÞ, where Z ¼ 0:02 is the structural loss factor. The constraints imposed on the RC
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 109

Table 1
Propeller-shafting system parameters

Parameter Value

mp (tonnes) 10
E s (GPa) 200
rs (tonnes/m3) 7.8
As (m2) 0.707
l s (m) 10.5
l se (m) 9
mb (tonnes) 0.2
kb (MN/m) 20000
cb (tonnes/s) 300

Table 2
Foundation parameters

Foundation parameter Value

a (mm) 1250
b (mm) 520
rf (kg/m3) 7700
E f (GPa) 200
uf 0.3
hf (mm) 10
f (deg) 15

Table 3
Hull parameters

Parameter Value

E h (GPa) 200
ro (m) 3.25
hh (mm) 45
l h (m) 60

107

106
Foundation four-pole parameters

105

104

103

102

101

100

10-1
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 5. Four-pole parameters of the foundation af .

 ARTICLE IN PRESS
110 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

Table 4
Optimisation limits and for the RC parameters

RC parameter Lower limit Upper limit

kr (MN/m) 15 1500
mr (tonnes) 1 20
cr (tonnes/s) 5 1100

Table 5
Optimal parameters for a single RC configuration

RC parameter opt(J T F ;W ) opt(J hPW i )

kr (MN/m) 168.51 145.95

mr (tonnes) 1 1
cr (tonnes/s) 287.43 67.313

parameters within the optimisation process are shown in Table 4. Minimisation of the maximum weighted
force transmissibility and the maximum weighted time-averaged power transmission to the hull have been
examined numerically in the following sections. The low-frequency range up to 100 Hz using a frequency
increment of 0.01 Hz has been considered. Results obtained using the method outlined by Goodwin [2] for
optimisation of the RC parameters are also presented for comparison.
Using a single RC configuration, the optimal values for the RC virtual mass, stiffness and damping
parameters are obtained by optimising the fitness criteria given by Eqs. (24) and (25). These two fitness criteria
correspond to minimising the maximum weighted force transmissibility, opt(J T F ;W ) and minimising the
maximum weighted time-averaged power transmission to the hull, opt(J hPW i ). The opt(J T F ;W ) and
opt(J hPW i ) parameter sets for a single RC configuration are given in Table 5. Fig. 6 shows the weighted
force transmissibility, T F ;W , versus frequency for the case without the RC, and using the opt(J T F ;W ) and
opt(J hPW i ) parameter sets. The largest peak in the response occurring at approximately 55 Hz can be
attributed to the fundamental propeller-shafting resonance, while the peaks at around 22, 46 and 73 Hz are
due to excitation of hull axial resonances. The introduction of the RC results in the elimination of the
dominant propeller-shafting resonance and introduces another resonance at approximately 16 Hz, as well as
significantly lowering the maximum response. It can be seen however that the opt(J hPW i ) parameter set does
not correlate to the optimum parameter set obtained when minimising the maximum weighted force
transmissibility, J T F ;W . A plot of the weighted time-averaged power transmission to the hull, hPW i, versus
frequency is presented in Fig. 7, again without the RC, and using the opt(J T F ;W ) and opt(J hPW i ) parameter
sets. This figure displays the same resonances as Fig. 6, however in this case the hull resonances are more
significant in amplitude. Similar to Fig. 6, Fig. 7 shows that the opt(J T F ;W ) parameter set does not result in the
minimal hPW i controlled response. It was observed that optimisation of T F ;W using the opt(J T F ;W ) parameter
set and hPW i using the opt(J hPW i ) parameter set using a single RC results in the two largest resonance peaks
in the controlled responses becoming equal. It was observed that a smaller virtual mass increases the effective
bandwidth of the response which is attenuated (that is, the suppression band). This result is contrary to the
traditional DVA which demonstrates a widening of the suppression band with an increase in the DVA mass.
The increase in the suppression band for a smaller virtual mass is due to the inertial force associated with the
RC being proportional to the relative motion of its two opposing terminals. Increasing the virtual mass
increases the impedance between the RC terminals. In the upper limit, an infinite virtual mass would reduce
the RC dynamically to a rigid link.
The opt(J T F ;W ) and opt(J hPW i ) parameter sets for a dual RC configuration consisting of two RCs in series
are given in Table 6. Figs. 8 and 9, respectively, present T F ;W and hPW i versus frequency for the dual RC
configuration, each showing the uncontrolled response and the controlled responses obtained using the two
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 111

101

100

TF,W
10-1

10-2

10-3
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 6. Log weighted force transmissibility of the propeller-shafting system: without the RC (– - -), with the RC tuned to minimise T F ;W
(––––) and with the RC tuned to minimise hPW i (- - - - -).

×1010
103

102

101

100
〈ΠW〉 (watts)

10-1

10-2

10-3

10-4

0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 7. Log weighted time-averaged power transmission to the propeller-shafting system: without the RC (– - -), with the RC tuned to
minimise T F;W (––––) and with the RC tuned to minimise hPW i (- - - - -).

Table 6
Optimal parameters for a dual RC configuration

RC parameter Opt(J T F ;W ) Opt(J hPW i )

kr;1 (MN/m) 116.91 93.192

mr;1 (tonnes) 1 1
cr;1 (tonnes/s) 573.32 71.610
kr;2 (MN/m) 596.73 1265.6
mr;2 (tonnes) 1.7859 3.1525
cr;2 (tonnes/s) 134.37 256.16
 ARTICLE IN PRESS
112 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

101

100

TF,W 10-1

10-2

10-3
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 8. Log weighted force transmissibility of the propeller-shafting system: without the RC (– - -), with 2 RCs tuned to minimise T F;W
(––––) and with 2 RCs tuned to minimise hPW i (- - - - -).

×1010
103

102

101

100
〈ΠW〉 (watts)

10-1

10-2

10-3

10-4

0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 9. Log weighted time-averaged power transmission to the propeller-shafting system: without the RC (– - -), with 2 RCs tuned to
minimise T F;W (––––) and with 2 RCs tuned to minimise hPW i (- - - - -).

fitness functions. These two figures show similar trends to Figs. 6 and 7 with the exception that three dominant
peaks in the optimal controlled responses have become equal as a result of the optimisation process. It is of
interest to note that one of the three equal peaks in the optimal controlled response for hPW i (Fig. 9)
corresponds to the hull resonance occurring at 73 Hz. This demonstrates the importance of including the
dynamic response of the hull when tuning the RC parameters.
A comparison of the controlled responses for T F ;W obtained using the Goodwin model (for one RC) and
the opt(J T F ;W ) parameter sets using the single and dual RC configurations is shown in Fig. 10. Similarly, a
comparison of the controlled responses for hPW i obtained using the Goodwin model (one RC) and the
opt(J hPW i ) parameter sets using one and two RCs in series is presented in Fig. 11. Both figures show that the
controlled response obtained using the dual RC configuration results in the lowest peak, while the Goodwin
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 113

101

100

TF,W
10-1

10-2

10-3
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 10. Log weighted force transmissibility of the propeller-shafting system: optimal RC using Goodwin’s method (– - -), with 1 RC tuned
to minimise T F;W (- - - - -) and with 2 RCs tuned to minimise T F ;W (––––).

×1010
103

102

101

100
〈ΠW〉 (watts)

10-1

10-2

10-3

10-4

0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 11. Log weighted time-averaged power transmission to the propeller-shafting system: optimal RC using Goodwin’s method (– - -),
with 1 RC tuned to minimise hPW i (- - - - -) and with 2 RCs tuned to minimise hPW i(––––).

parameter set results in the worst controlled response over the frequency range of interest. This information is
presented graphically in Fig. 12 in terms of a percentage reduction of the maximum response (relative to the
uncontrolled response in the absence of a RC) for all the parameter sets. Fig. 12 clearly shows that the
opt(J hPW i ) parameter sets for both one or two RC configurations achieves significant reduction in hPW i, but
has a poor T F ;W response compared with the other parameter sets. In contrast, the opt(J T F ;W ) parameter sets
using one and two RCs results in excellent reduction of both the T F ;W and hPW i responses.
It is of interest to perform a perturbation analysis in order to observe the effect of any variations in the
system parameters from an idealised case. A general procedure for studying the parameter sensitivities could
be used to assess the change in the system’s response to variations in the RC parameters. However due to the
sufficiently large modal spacing, discrete variations on the RC parameters of 10% is regarded to be adequate
for the perturbation analysis considered within this work. In this investigation, perturbations of the thrust
 ARTICLE IN PRESS
114 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

100.0

99.5

99.0
Percentage reduction
98.5

98.0

97.5

97.0

96.5

96.0

95.5

95.0
Goodwin opt(JTF,W) opt(JTF,W) opt(J〈ΠW〉) opt(J〈ΠW〉)
1RC 1 RC 2 RCs 1 RC 2 RCs

Fig. 12. Percent reduction in the response for the various parameter sets relative to the response without the RC: T F;W , hPW i.

101

100
TF,W

10-1

10-2

10-3
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 13. Log weighted force transmissibility of the propeller-shafting system: with 2 RCs tuned to minimise T F;W (––––) and with 2
perturbed RCs tuned to minimise T F ;W (- - - - -).

bearing and RC parameters have been considered. The thrust bearing was chosen since its dynamic properties
are nonlinear and vary considerably within its operating range [1]. It was found however that major
perturbations provided insignificant changes in both the T F ;W and hPW i responses. This is due to the large
stiffness and low mass of the thrust bearing relative to the other propeller-shafting system components. Figs.
13 and 14, respectively, show the controlled responses for T F ;W and hPW i using two RCs in series. In each
figure, the optimal controlled response is shown as well as the controlled response obtained by perturbing the
optimal RC parameters by 10%. All combinations of the perturbation quantities (by 10%) on the RC
parameters were considered and the combination which resulted in the worst-case scenario (largest increase in
 ARTICLE IN PRESS
P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116 115

×1010
103

102

101

100
〈ΠW〉 (watts)
10-1

10-2

10-3

10-4

0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)

Fig. 14. Log weighted time-averaged power transmission to the propeller-shafting system: with 2 RCs tuned to minimise hPW i (––––) and
with 2 perturbed RCs tuned to minimise hPW i (- - - - -).

Table 7
Percentage perturbations on RC parameters

RC parameter opt(J T F;W ) (%) opt(J hPW i ) (%)

kr;1 (MN/m) 10 +10

mr;1 (tonnes) +10 +10
cr;1 (tonnes/s) +10 10
kr;2 (MN/m) 10 +10
mr;2 (tonnes) +10 10
cr;2 (tonnes/s) +10 10

the controlled response) is given in Table 7. Results showed that perturbations in the RC parameters did not
significantly affect either the T F ;W and hPW i controlled responses.

5. Conclusions

The dynamic response of the propeller-shafting system in a submarine has been modelled using a modular
transmission matrix description allowing for flexibility in future design modifications. Both propeller-shafting
and hull resonances were present when examining the force transmissibility and power transmission to the
hull, which demonstrates the importance of including the hull in the dynamic modelling. Optimal RC
parameters have been obtained by minimising the frequency weighted maximum force transmissibility and
time-averaged power transmission to the submarine hull over a specified frequency range. An optimisation
scheme using a genetic and a general nonlinear constrained algorithm was applied to the fitness criteria for
single and dual RC configurations. Realistic lower and upper bounds on the RC parameters were applied as
constraints within the optimisation process. It was shown that minimisation of the power transmission does
not necessarily provide the minimum force transmissibility. Perturbations of the thrust bearing and resonance
changer parameters were shown to have no significant effect on the controlled responses.
 ARTICLE IN PRESS
116 P.G. Dylejko et al. / Journal of Sound and Vibration 300 (2007) 101–116

References

[1] J. Pan, N. Farag, T. Lin, R. Juniper, Propeller induced structural vibration through the thrust bearing, Proceedings of Acoustics 2002,
Adelaide, Australia, 13–15 November 2002, pp. 390–399.
[2] A.J.H. Goodwin, The design of a resonance changer to overcome excessive axial vibration of propeller shafting, Institute of Marine
Engineers—Transactions 72 (1960) 37–63.
[3] D.W. Lewis, P.E. Allaire, P.W. Thomas, Active magnetic control of oscillatory axial shaft vibrations in ship shaft transmission
systems, part 1: system natural frequencies and laboratory scale model, Tribology Transactions 32 (1989) 170–178.
[4] H. Frahm, 1911 US Patent 989,958, Device for damping vibrations of bodies.
[5] J.B. Hunt, Dynamic Vibration Absorbers, Mechanical Engineering Publications LTD, London, 1979.
[6] B.G. Korenev, L.M. Reznikov, Dynamic Vibration Absorbers Theory and Technical Applications, Wiley, Chichester, 1993.
[7] D. Hartog, Mechanical Vibrations, McGraw-Hill, New York, 1956.
[8] V.H. Neubert, Dynamic absorbers applied to a bar that has solid damping, Journal of the Acoustical Society of America 36 (1964)
673–680.
[9] L. Kitis, B.P. Wang, W.D. Pilkey, Vibration reduction over a frequency range, Journal of Sound and Vibration 89 (1983) 559–569.
[10] A.H.P. Lau, G.C.K. Lam, The optimal design of dynamic absorber in the time domain and the frequency domain, Journal of Applied
Acoustics 28 (1989) 67–78.
[11] H.J. Rice, Design of multiple vibration absorber systems using modal data, Journal of Sound and Vibration 160 (1993) 378–385.
[12] D.A. Rade, V.J. Steffen, Optimisation of dynamic vibration absorbers over a frequency band, Journal of Sound and Vibration 14
(2000) 679–690.
[13] J.C. Snowdon, Mechanical four-pole parameters and their application, Journal of Sound and Vibration 15 (1971) 307–323.
[14] S. Rubin, Mechanical immittance and transmission-matrix concepts, Journal of the Acoustical Society of America 41 (1966)
1171–1179.
[15] R.E.D. Bishop, D.C. Johnson, The Mechanics of Vibration, Cambridge University Press, Cambridge, 1960.
[16] W.C.L. Hu, D.D. Kana, Four-pole parameters for impedance analysis of conical and cylindrical shells under axial excitations,
Journal of the Acoustical Society of America 43 (1968) 683–690.
[17] E. Skudrzyk, Simple and Complex Vibratory Systems, The Pennsylvania State University Press, London, 1968.
[18] A. Osyczka, Evolutionary Algorithms for Single and Multicriteria Design Optimization, Physica-Verlag, New York, 2002.
[19] D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning, Addison-Wesley, Reading, MA, 1989.
[20] F. Erbatur, I. Hassancebi, I. Tutuncu, H. Kilic, Optimal design of planar and space structures with genetic algorithms, Computers and
Structures 75 (2000) 209–224.